draineye/draineye-defects
收藏Hugging Face2026-03-20 更新2026-03-29 收录
下载链接：
https://hf-mirror.com/datasets/draineye/draineye-defects
下载链接
链接失效反馈官方服务：
资源简介：
---
license: cc-by-nc-4.0
task_categories:
- image-classification
language:
- en
size_categories:
- 1K<n<10K
---
# DrainEye — Drainage Defect Classification Dataset
A balanced image classification dataset for detecting defects in drainage
systems, assembled from 5 public Kaggle sources.
## Classes
- **corrosion** — pipe and surface corrosion
- **crack** — concrete and pavement cracks
- **no_damage** — clean, undamaged surfaces
## Dataset Structure
| Split | Images | Share |
|-------|--------|-------|
| train | 1050 | 70% |
| val | 225 | 15% |
| test | 225 | 15% |
| **Total** | **1500** | |
500 images per class, balanced with `random.seed=42`.
## Sources
- [Surface Crack Detection](https://www.kaggle.com/datasets/arunrk7/surface-crack-detection)
- [Concrete and Pavement Crack Images](https://www.kaggle.com/datasets/oluwaseunad/concrete-and-pavement-crack-images)
- [Pipeline Corrosion Dataset](https://www.kaggle.com/datasets/aditya068/pipeline-corrosion-dataset)
- [Corrosion Detection](https://www.kaggle.com/datasets/bsurya27/corrosion-detection)
- [Corrosion Detect Dataset](https://www.kaggle.com/datasets/wednesday233/corrosion-detect-dataset)
## Intended Use
Baseline training for a mobile drainage inspection app (Flutter + TFLite).
## Loading
```python
from datasets import load_dataset
dataset = load_dataset("draineye/draineye-defects")
```
提供机构：
draineye
